Step Functions¶

Overview¶

• Model workflows as state machines.
• Serverless workflows to orchestrate lamba functions.
• Support sequence, parallel, conditions, timeouts, error handling.
• Integrates with EC2, EC2, On-Prem Servers, API gateway etc.
• Maximum execution time is 1 year.
• Supports human/manual approval feature.
• Useful for order fulfillment, data processing (ETL), web applications (payment processing).

States¶

Error Handling¶

The default failure behaviour is to fail the entire execution if a state reports an error.

An error can occur due to:

• Task failures.
• Machine definition issues (no matching rule etc).
• Transient issues (network partition event).

Try to include data in the error messages to help with troubleshooting.

Pre-defined Error Codes¶

There's four pre-defined error codes that can be returned -

Retrying¶

Applies to tasks, or parallel state. A task can use multiple retry methods that are evaluated in order.

There's several parameters that can be used to modify how often a retry is attempted, how many times to retry before giving up, and a back off rate to increase the interval between each subsequent retry.

``` json
"HelloWorld": {
  "Type": "Task",
  "Resource": "arn:aws:lambda:REGION:ACCOUNT_ID:function:FUNCTION_NAME",
  "Retry": [
    {
      "ErrorEquals": ["CustomError"],
      "IntervalSeconds": 1,
      "MaxAttempts": 2,
      "BackoffRate": 2.0
    },
    {
      "ErrorEquals": ["States.TaskFailed"],
      "IntervalSeconds": 30,
      "MaxAttempts": 2,
      "BackoffRate": 2.0
    },
    {
      "ErrorEquals": ["States.ALL"],
      "IntervalSeconds": 5,
      "MaxAttempts": 5,
      "BackoffRate": 2.0
    }
  ],
  "End": true
}
```

Catching Errors¶

Applies to Tasks or parallel steps, and has two attributes to control what errors to catch, what the next step should be, and what to include in the result

``` json
"HelloWorld": {
  "Type": "Task",
  "Resource": "arn:aws:lambda:REGION:ACCOUNT_ID:function:FUNCTION_NAME",
  "Catch": [
    {
      "ErrorEquals": ["CustomError"],
      "Next": "CustomErrorCallback"
    },
    {
      "ErrorEquals": ["States.TaskFailed"],
      "Next": "ReservedTypeCallback"
    },
    {
      "ErrorEquals": ["States.ALL"],
      "Next": "NextTask",
      "ResultPath": "$.error"
    }
  ],
  "End": true
},
"NextTask": {
  "Type": "Pass",
  "Result": "This is a fallback from a reserved error code.",
  "End": true
},
"CustomErrorFallback": {
  "Type": "Pass",
  "Result": "This is a fallback from a custom lambda function exception.",
  "End": true
}
```

Standard Step Functions¶

• Max duration of 1year.
• Execution start rate up to 2,000/sec.
• State transition rate over 4,000/sec per account.
• Priced per state transition.
• Can be listed and described with step function APIs, and visually debugged through the console.
• Can be inspected in CloudWatch Logs by enabling logging on the state machine.
• Exactly-once workflow execution.

Express Step Functions¶

• Max duration of 5mins.
• Execution start rate up to 100,000/sec.
• State transition rate is unlimited.
• Prices by number of executions run, their duration, and memory consumption (cheaper).
• Can be inspected in CloudWatch Logs by enabling logging on the state machine.
• At-least-once workflow execution.